Role of media coverage in mitigating COVID-19 transmission: Evidence from China.

This paper evaluates the impact of COVID-19 media coverage in mitigating its spread in China during the early phase of the pandemic. We construct a provincial-level dataset on COVID-19 and link it with population mobility data, among other control variables, to estimate how media coverage mitigates the spread of COVID-19. Seemingly unrelated regressions are used to examine the simultaneous impact of media coverage on the number of new cases and close contacts. The results show that the effect of media coverage on COVID-19 transmission in China had an inverse-U curvature and was mediated by within- and across-province population mobility. Our simulation results indicate that COVID-19 media coverage in China was associated with a potential reduction of 394,000 cases and 1.4 million close contacts during January 19 and February 29, 2020. Our results also provide strong support for the use of contact tracing in mitigating COVID-19 transmission.

Introduction

Effective implementation of government interventions and policies to prevent and control an ongoing pandemic relies on the support for, compliance to, and trust of the policies among the general public (Saksena, 2018). The course of a pandemic is determined by individual and collective actions of people, who internalize the information available to them (Gersovitz and Hammer, 2003). Thus, media coverage of an ongoing pandemic may play a crucial role in mitigating the spread of the pandemic (Islam et al., 2020). Information about the severity, mortality, and modes of transmission of the disease available to the public improves the compliance to government policies and directives (Gersovitz and Hammer, 2003).

COVID-19, a disease caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), was first reported in China on December 30, 2019. It has since spread outside of China and was declared a worldwide pandemic on March 11, 2020. By July 9, 2020, China reported 85,399 cases of COVID-19 and 4,648 associated deaths (Guan et al., 2020), while the global case count stood at 30,675,675 as of September 20, 2020 (World Health Organization, 2020). The cluster of mysterious pneumonia cases was first reported in Wuhan, a megacity with a population of 11 million in Hubei province (Li et al., 2020). Chinese central and local governments took extraordinary measures to implement a wide range of interventions and policies to control the spread of COVID-19, including contact tracing, identifying the causative pathogen, genomic characterization of the pathogen, developing testing kits, mandating the use of facemasks, and social distancing (Chen et al., 2020a; Zhu et al., 2020). On January 20, China activated the highest level of public health emergency mobilization across all sectors in response to the COVID-19 epidemic (Figure 1 ). The City of Wuhan was shutdown to limit mobility starting on January 23. The declaration of a state of emergency has been shown to reduce mobility during the COVID-19 pandemic (Katafuchi et al., 2020). In late February 2020, the exponential growth of the number of confirmed cases in China was tamped down (Maier and Brockmann, 2020).

Figure 1

Public Health Emergency Responses by Province in China. Note: In China, public health emergencies, including infectious disease epidemics, can be categorized into four levels, with Level I being the highest level of mobilization.

The prevention and control of COVID-19 in China is challenging. Wuhan is a crucial transportation hub in central China with connecting railway and flight networks, heightening the risks of COVID-19 transmission (Nakamura and Managi, 2020). The Lunar New Year Holiday, January 24 to 30, in 2020, is one of the most celebrated national holidays in China, typically with more than 0.45 billion travelers in January and early February (Tian et al., 2020). The intense population mobility associated with Wuhan and the holiday season, coupled with a completely new disease with many features unbeknownst to the scientists even many months later, has posed a challenge to the Chinese authorities with profound consequences. Given China had experienced a similar but smaller-scale epidemic of SARS in 2003, there is a debate about if, when, and how information availability and media coverage have mitigated the spread of the pandemic.

Media coverage has a crucial role in disseminating and advocating public policies and information when emergencies occur and securing the public's attention, support, and compliance (Degeling and Kerridge, 2013; Otten, 1992). The emergency of a new infectious disease might lead to confusion and panic if no proper information was available in time. For example, compliance with the home isolation policy had been an issue in Israel when the public was not well informed on home isolation policies and guidelines (Dickens et al., 2020). Media coverage has been examined in political science, finance, and health (Boukes et al., 2015; Cieslak and Schrimpf, 2019; Jarlenski and Barry, 2013; Kasper et al., 2015). In public health, communication is key to disseminating information related to diseases and interventions, such as tobacco control (Smith et al., 2008), mental illness (Wahl, 2003), obesity (Niederdeppe and Frosch, 2009), and infectious disease (Degeling and Kerridge, 2013; Saksena, 2018). Although there are debates that news report may be influenced by political considerations (Hayes et al., 2007; Saksena, 2018), and how to ‘frame’ the events may have unintended consequences (Jarlenski and Barry, 2013; Kostadinova and Dimitrova, 2012), news is still the primary, if imperfect, source of information for most people on public issues and debates (Jarlenski and Barry, 2013).

In this paper, we estimate the effects of media coverage on COVID-19 prevention and control in China. Following the Standard Inflammatory Response (SIR) model (or susceptible-infected-recovered model as referred to elsewhere) to investigate pandemic transmission outlined in Adda (2016), we model the within- and across-province spread of COVID-19 and the effects of provincial-specific media coverage using provincial-level data. We use the daily number of new cases and close contacts at the provincial-level to describe the temporal and spatial spread of COVID-19 and the daily accumulated number of official news reports on COVID-19 in every province to proxy provincial COVID-19 media coverage. We evaluate the impact of media coverage by simulating the counterfactual when media coverage was absent.

The remainder of this paper proceeds as follows. Section 2 describes the data. Section 3 outlines the econometric method used in the paper. Section 4 presents the study results. Section 5 describes the counterfactual simulation. Section 6 discusses the implications and limitations of the study, and Section 7 concludes.

Data

We compiled data on COVID-19, media coverage, population mobility, and control variables from various sources. The official data on COVID-19 were available since January 20, 2020 for Chinese provinces except for Hubei, for which the data can be dated back to January 1, 2020, see, e.g., Tian et al. (2020). Several provinces had lowered the level of emergency response, as shown in Figure 1, and gradually reopened in late February. Therefore, we chose the end of our study period as February 29.

COVID-19 data

We extracted the number of daily new COVID-19 cases and the number of daily identified close contacts for the 31 provincial administrative units in mainland China from the websites of central and provincial health authorities. Most studies on COVID-19 from China have used the daily number of confirmed cases as the major indicator of interest (Pan et al., 2020; Qiu et al., 2020; Tian et al., 2020). We also examined the number of close contacts1 because it is a crucial alternative measure of the spread of COVID-19, among which some later confirmed to have COVID-19 infection. Successful prevention and control of the pandemic often involve intensive efforts in contact tracing, i.e., identifying close contacts of the confirmed cases, and appropriate follow-up measures, including self-isolation or quarantining of the close contacts (Maier and Brockmann, 2020). Therefore, we use both indicators to examine the temporal and spatial spread of COVID-19 (see the temporal changes in Figure 2 ).

Figure 2

Daily Numbers of COVID-19 New Cases and Close Contacts in Mainland China

Media coverage

We collected the official news releases and reports on COVID-19 for each province to measure media coverage. We used the cumulative daily number of news reports and releases (# news) to measure the intensity of media coverage. The news reports were extracted by using Python from DXY Inc, a leading Chinese digital service provider and news synthesizing platform in the healthcare sector. DXY built an information portal for COVID-19 in early January, which had 41.6 billion visits by July 14, 2020.2

The first news release appeared on December 31, 2019, reporting 27 cases of pneumonia of unknown etiology in Wuhan. During our study period, there were a total of 7,321 official news releases and reports about COVID-19. We included both local news and reports released by the 31 provinces and the reports and news released at the national-level but relevant to a specific province. We constructed a final set of 1,849 news for 31 provinces (see Appendix A). Instead of content analysis, we calculated the cumulative number of news releases or reports for each province each day and then calculated the cumulative daily number of news to measure the extent of media coverage, as we have explained earlier. A detailed description of DXY data and the collection and measurement of the news releases and reports is in Appendix A.

We chose to use official news reports from major news outlets and national and provincial health authorities' websites. News reports on the pandemic abound but authoritative information was limited in the early stage of the pandemic (Degeling and Kerridge, 2013; Saksena, 2018). Official news releases and reports presented trustworthy information with impact and accountability, led to concerted public responses, and helped to set public policy agendas (Jarlenski and Barry, 2013). Other news sources had often used and adapted those reports.

We used the daily cumulated number of news releases and reports as a key measure of media coverage. The information on COVID-19, particularly the scientific findings and prevention and control policies, had been continuously developing and adapting, posing difficulties for content analysis. Thus, instead of the content analysis often used in communication studies, we chose to use the daily cumulated number of official news reports to measure media coverage.

Population mobility

The population mobility indicators from Baidu Inc. included the index of population inflow across provinces and the index of within-province population movement into the provincial capital cities.3 Baidu launched its product “Baidu Mobility (Baidu Qianxi in Chinese)” in 2014, which illustrates daily population inflow for each province using mapping tools and information technologies, including Location-based Services. The plots of population inflow and movement in every province were shown in Appendix B (Appendix Figure B1 and Figure B2). The Baidu mobility data and similar data from Tencent had been used in COVID-19 research in China elsewhere (Qiu et al., 2020; Tian et al., 2020).

Figure B1

Plots of the variation for population inflow and movement in each province

Figure B2

Plots of the variation for population inflow for each province in 2019 and 2020

Control variables

Our control variables included provincial-level weather data, area, and inter-province distance indicated by the distance between capital cities. The data on area and distance were collected and calculated from the 2018 China Statistical Yearbook and the China Land & Resources Almanac.

Weather and temperature may affect the life span and transmission of SARS-COV-2 (Lin et al., 2006), through both the direct effect on the virus and the indirect effect through behavioral changes related to social gatherings (Adda, 2016). We used daily average temperature, wind, and precipitation of the capital city for every province to indicate the daily weather as in Qiu et al. (2020). The weather data was collected from the National Meteorological Center of China Meteorological Administration (http://www.nmc.cn/).

Earlier studies used indicators, including the provincial per-capita GDP, as mediating socioeconomic factors (Qiu et al., 2020; Tian et al., 2020). Our provincial fixed-effects capture the provincial-level socioeconomic conditions.

Appendix C provides the description and summary statistics of key variables.

Econometric specification

To explore the impacts of economic activity on the spread of infectious disease, Adda (2016) developed a within- and across-province model (henceforth Adda model). Our model extended the SIR model and described a more comprehensive alternative, in which the spread of infectious disease depends on the local number of cases and population inflow.

We estimated the effects of media coverage with lags of 3-, 5-, and 7-days to model the impact of different incubation periods because the reported incubation period of SARS-CoV-2 is about 5.2 days (Guan et al., 2020; Li et al., 2020). We also estimated the potential effects of media coverage on COVID-19 prevention and control through reduced within- and across-province population mobility.

The within-province model

We began our estimation by the traditional within-province model as presented in the Seemingly Unrelated Regression (SUR) system, Eq. (1) and (2), to explore the spread of COVID-19. and are the logarithmic transformation of daily new cases and close contacts in the province on day . is the susceptible population, and is incubation time. The lagged is also in the logarithmic form. indicates control variables, including provincial and date fixed effects. To test the potential variation in the incubation period, we set as 3, 5, and 7 days. We also use Eqs. (3) and (4) to explore the association between daily new cases and close contacts. The calculation of the susceptible population is a challenge. There was no vaccine for COVID-19 during the study period, thus anyone could be infected – although age may be a factor because 87% of the patients aged between 30-79 (Wu and McGoogan, 2020). We chose to use the whole provincial population to proxy the susceptible population but recognize the limitations in doing so. Studies suggested a portion of the populations may be less likely to have COVID-19 because of prior infections of the common strains of coronavirus. However, if the proportion does not vary significantly across the provinces, which seems to be the case, our use of the provincial population only changes the scale of the coefficient.

The basic across-province model

Eqs. (3) and (4) predict that the spread of virus and disease in each province may be affected by the within- and across-province transmission. , and are also the logarithm form of the counts, and is the provinces other than . indicates control variables including provincial and daily fixed effects and the full control of land areas in the province, inter-province distances, and weather conditions. To capture the differences in closeness across provinces, we weight by the inverse of the distance between the two provinces. The same transformation of new patients and close contacts is estimated by Eqs. (7) and (8). Identified new COVID-19 cases and close contacts would be quarantined or under medical observation in the province where their condition or status was ascertained, so we chose not to include the across-province item for and in the right side of Eqs. (7) and (8), and the same treatment is used in the following estimations.

The full across-province model

The Eqs. (9)-(12) follows Eqs. (5)-(8) where the spread of the disease may be determined by both within- and across-province factors. and are lagged province-specific variable vectors (with dimensions being and ) that may influence the spread of disease within- and across-province.

Separate regressions: Before and after February 5

We run separate regressions for the full sample (T) and two subsamples, i.e., the sample with data from January 1 to February 5 (T) and the sample with data from February 6 to February 29 (T), as the national number of new confirmed cases peaked on February 5. We intended to examine the difference in the patterns before and after the peak.

Robustness check: Excluding Hubei province

Data for Hubei province were amended on April 16, with 325 cases added due to earlier omissions or misreporting. However, there was no information as to which dates the added cases occurred. In addition, the majority of the cases occurred in Hubei. Therefore, as a robustness check, we run additional estimations using the sample excluding data from Hubei.

Results

Baseline models

Table 1 presents the results for estimating Eqs. (1)-(4). When the lag is set at 3 and the full time period (T) is used, a 100% change in the number of new cases is associated with 24% increase in the number of news cases three days later,4 and a 100% increase in the number of close contacts is associated with an increase of 27% in the number of close contacts 3 days later. After adding the current period of close contacts and new cases as explanatory variables in the SUR estimation, a 100% increase in the number of close contacts is associated with an increase of 26% in the number of new cases, while a 100% increase in the confirmed case leads to an increase of 109% in the number of close contacts. Separate regressions for the samples before and after February 5 suggest that the effects are stronger in t and reduced in t. The impact also decreased as the lag increases from 3 days to 7 days, except for the association between daily new cases and close contacts, which has strengthened across the models with the lag of 3-, 5- to 7 days.

Table 1

Daily spread of COVID−19 within province

	T		T1		T2		T		T1		T2
	R	C	R	C	R	C	R	C	R	C	R	C
Panel A: 3 days lag
ɑwithin,	0.24⁎⁎⁎	0.27⁎⁎⁎	0.30⁎⁎⁎	0.33⁎⁎⁎	0.10⁎⁎⁎	0.17⁎⁎⁎	0.17⁎⁎⁎	0.01	0.15⁎⁎⁎	−0.08⁎⁎⁎	0.08⁎⁎⁎	0.12⁎⁎⁎
	(0.01)	(0.02)	(0.01)	(0.03)	(0.02)	(0.03)	(0.01)	(0.02)	(0.01)	(0.03)	(0.02)	(0.03)
Close contacts							0.26⁎⁎⁎		0.44⁎⁎⁎		0.12⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.09⁎⁎⁎		1.37⁎⁎⁎		0.47⁎⁎⁎
								(0.05)		(0.04)		(0.07)
N	1767	1767	961	961	806	806	1767	1767	961	961	806	806
Panel B: 5 days lag
ɑwithin	0.20⁎⁎⁎	0.26⁎⁎⁎	0.25⁎⁎⁎	0.28⁎⁎⁎	0.06⁎⁎⁎	0.23⁎⁎⁎	0.13⁎⁎⁎	0.05*	0.11⁎⁎⁎	−0.06⁎⁎	0.03⁎⁎	0.20⁎⁎⁎
	(0.01)	(0.02)	(0.02)	(0.03)	(0.02)	(0.03)	(0.01)	(0.02)	(0.02)	(0.03)	(0.02)	(0.03)
Close contacts							0.27⁎⁎⁎		0.47⁎⁎⁎		0.13⁎⁎⁎
							(0.01)		(0.02)		(0.02)
New cases								1.08⁎⁎⁎		1.37⁎⁎⁎		0.49⁎⁎⁎
								(0.04)		(0.04)		(0.07)
N	1705	1705	899	899	806	806	1705	1705	899	899	806	806
Panel C: 7 days lag
ɑwithin	0.16⁎⁎⁎	0.18⁎⁎⁎	0.24⁎⁎⁎	0.21⁎⁎⁎	0.03*	0.12⁎⁎⁎	0.11⁎⁎⁎	−0.01	0.15⁎⁎⁎	−0.14⁎⁎⁎	0.01	0.10⁎⁎⁎
	(0.01)	(0.02)	(0.02)	(0.03)	(0.02)	(0.04)	(0.01)	(0.02)	(0.02)	(0.03)	(0.02)	(0.04)
Close contacts							0.28⁎⁎⁎		0.47⁎⁎⁎		0.15⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.14⁎⁎⁎		1.43⁎⁎⁎		0.57⁎⁎⁎
								(0.04)		(0.05)		(0.07)
N	1643	1643	837	837	806	806	1643	1643	837	837	806	806
Province FE	√	√	√	√	√	√	√	√	√	√	√	√
Date FE	√	√	√	√	√	√	√	√	√	√	√	√
Controls	×	×	×	×	×	×	×	×	×	×	˟	˟

Notes: 1. Standard errors in parentheses; 2. * p<0.1,

p<0.05,

p<0.01; 3. T=the full sample, T=subsample with data from Jan 1 to Feb 5, T=subsample with data from Feb 6 to Feb 29; 4. R=new cases, C=Close contacts.

Results of Eqs. (5)-(8) are in Table 2 . After adding the inter-province correlation , results for the within effect and the association between the number of close contacts and new cases only have trivial changes. Across effect () is only statistically significant for the whole time period (T) and after February 5 (T2). The across-effect of new cases is positive in the models with a 5-day lag, consistent with the conjecture that the incidence in one province generates additional cases in other provinces (Adda, 2016). However, the across-effect for the number of close contacts are difficult to interpret for the models with 3- and 5-days lag.

Table 2

Daily spread of COVID−19 across provinces

	T		T1		T2		T		T1		T2
	R	C	R	C	R	C	R	C	R	C	R	C
Panel A: 3 days lag
ɑwithin	0.23⁎⁎⁎	0.23⁎⁎⁎	0.29⁎⁎⁎	0.34⁎⁎⁎	0.10⁎⁎⁎	0.13⁎⁎⁎	0.17⁎⁎⁎	−0.03	0.14⁎⁎⁎	−0.06	0.08⁎⁎⁎	0.09⁎⁎
	(0.01)	(0.03)	(0.02)	(0.04)	(0.02)	(0.03)	(0.01)	(0.03)	(0.02)	(0.04)	(0.02)	(0.03)
ɑacros	−0.01	−0.04⁎⁎	−0.01	0.01	−0.00	−0.04⁎⁎	0.00	−0.03*	−0.01	0.02	0.00	−0.04⁎⁎
	(0.01)	(0.02)	(0.02)	(0.03)	(0.01)	(0.02)	(0.01)	(0.02)	(0.02)	(0.03)	(0.01)	(0.02)
							0.26⁎⁎⁎		0.44⁎⁎⁎		0.12⁎⁎⁎
Close contacts							(0.01)		(0.01)		(0.02)
								1.09⁎⁎⁎		1.37⁎⁎⁎		0.47⁎⁎⁎
New cases								(0.05)		(0.04)		(0.07)
N	1767	1767	961	961	806	806	1767	1767	961	961	806	806
Panel B: 5 days lag
ɑwithin	0.20⁎⁎⁎	0.27⁎⁎⁎	0.24⁎⁎⁎	0.29⁎⁎⁎	0.05	−0.15	0.13⁎⁎⁎	0.05⁎⁎	0.10⁎⁎⁎	−0.04	0.07	−0.18
	(0.01)	(0.02)	(0.02)	(0.04)	(0.09)	(0.18)	(0.01)	(0.02)	(0.02)	(0.03)	(0.09)	(0.18)
ɑacros	0.01⁎⁎⁎	0.01⁎⁎⁎	−0.01	0.00	−0.02	−0.38⁎⁎	0.00⁎⁎⁎	0.01⁎⁎	−0.01	0.02	0.04	−0.38⁎⁎
	(0.00)	(0.00)	(0.02)	(0.03)	(0.09)	(0.18)	(0.00)	(0.00)	(0.01)	(0.02)	(0.09)	(0.18)
Close contacts							0.27⁎⁎⁎		0.47⁎⁎⁎		0.13⁎⁎⁎
							(0.01)		(0.02)		(0.02)
New cases								1.08⁎⁎⁎		1.37⁎⁎⁎		0.49⁎⁎⁎
								(0.04)		(0.04)		(0.07)
N	1705	1705	899	899	806	806	1705	1705	899	899	806	806
Panel C: 7 days lag
ɑwithin	0.16⁎⁎⁎	0.20⁎⁎⁎	0.24⁎⁎⁎	0.19⁎⁎⁎	0.03*	0.15⁎⁎⁎	0.10⁎⁎⁎	0.02	0.15⁎⁎⁎	−0.15⁎⁎⁎	0.01	0.14⁎⁎⁎
	(0.02)	(0.03)	(0.02)	(0.04)	(0.02)	(0.04)	(0.01)	(0.03)	(0.02)	(0.03)	(0.02)	(0.04)
ɑacros	−0.00	0.03	−0.01	−0.01	0.00	0.03*	−0.01	0.03*	−0.00	−0.00	−0.00	0.03*
	(0.01)	(0.02)	(0.01)	(0.02)	(0.01)	(0.02)	(0.01)	(0.02)	(0.01)	(0.01)	(0.01)	(0.02)
Close contacts							0.28⁎⁎⁎		0.47⁎⁎⁎		0.15⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.14⁎⁎⁎		1.43⁎⁎⁎		0.57⁎⁎⁎
								(0.04)		(0.05)		(0.07)
N	1643	1643	837	837	806	806	1643	1643	837	837	806	806
Province FE	√	√	√	√	√	√	√	√	√	√	√	√
Date FE	√	√	√	√	√	√	√	√	√	√	√	√
Controls	×	×	×	×	×	×	×	×	×	×	˟	˟

Notes: 1. Standard errors in parentheses; 2. * p<0.1

p<0.05

p<0.01; 3. T=the full sample, T=subsample with data from Jan 1 to Feb 5, T=subsample with data from Feb 6 to Feb 29; 4. R=new cases, C=Close contacts.

The impact of media coverage

We estimate Eqs. (9) and (12) with a set of variables on media coverage. We include a quadratic term of the number of news reports as media coverage may have a nonlinear impact on the spread of the pandemic. Media coverage might increase as the number of cases grew, but at the later stage of the epidemic, the cumulative impact of new coverage will manifest and limit the spread of the disease through reduced mobility and increased adherence to social distancing and other mitigation measures.

The estimation results are presented in Table 3 . Media coverage has a limited impact on the spread of this epidemic in the early stage, but the impact grew stronger after February 5 (T2). Media coverage in other provinces have statistically significant but small effects on the number of close contacts for the models with 3 and 5-day lags. The magnitude of the impact of media coverage decreased as the lag increased from 3, 5, to 7-days. The introduction of media coverage has only trivial changes on the association between the number of new cases and the number of close contacts relative to the baseline models.

Table 3

The effects of media coverage on daily spread of COVID−19 across provinces

	T		T1		T2		T		T1		T2
	R	C	R	C	R	C	R	C	R	C	R	C
Panel A: 3 days lag
# news, within	1.04⁎⁎⁎	0.79⁎⁎⁎	0.20⁎⁎	0.47⁎⁎⁎	1.23⁎⁎⁎	0.77⁎⁎⁎	0.84⁎⁎⁎	−0.37⁎⁎⁎	−0.01	0.20	1.16⁎⁎⁎	0.29
	(0.05)	(0.11)	(0.09)	(0.16)	(0.08)	(0.17)	(0.05)	(0.12)	(0.08)	(0.14)	(0.08)	(0.20)
# news2, within	−0.67⁎⁎⁎	−0.48⁎⁎⁎	0.04	−0.17	−0.84⁎⁎⁎	−0.49⁎⁎⁎	−0.55⁎⁎⁎	0.27⁎⁎⁎	0.11*	−0.22*	−0.80⁎⁎⁎	−0.17
	(0.04)	(0.08)	(0.07)	(0.13)	(0.06)	(0.12)	(0.04)	(0.09)	(0.06)	(0.11)	(0.06)	(0.14)
# news, across	−0.00	0.00	−0.01	−0.02⁎⁎	−0.00	−0.02⁎⁎	0.00	−0.02*	−0.01	0.02	0.00	−0.02⁎⁎
	(0.00)	(0.02)	(0.01)	(0.01)	(0.00)	(0.01)	(0.00)	(0.01)	(0.01)	(0.02)	(0.00)	(0.01)
Close contacts							0.25⁎⁎⁎		0.44⁎⁎⁎		0.08⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.12⁎⁎⁎		1.39⁎⁎⁎		0.39⁎⁎⁎
								(0.05)		(0.05)		(0.08)
N	1767	1767	961	961	806	806	1767	1767	961	961	806	806
Panel B: 5 days lag
# news, within	0.85⁎⁎⁎	0.56⁎⁎⁎	0.03	0.47⁎⁎⁎	1.02⁎⁎⁎	0.70⁎⁎⁎	0.70⁎⁎⁎	−0.39⁎⁎⁎	−0.20⁎⁎	0.44⁎⁎⁎	0.94⁎⁎⁎	0.23
	(0.06)	(0.13)	(0.10)	(0.18)	(0.09)	(0.18)	(0.06)	(0.13)	(0.09)	(0.15)	(0.09)	(0.20)
# news2, within	−0.55⁎⁎⁎	−0.37⁎⁎⁎	0.15*	−0.20	−0.71⁎⁎⁎	−0.47⁎⁎⁎	−0.46⁎⁎⁎	0.25⁎⁎⁎	0.24⁎⁎⁎	−0.41⁎⁎⁎	−0.66⁎⁎⁎	−0.14
	(0.04)	(0.09)	(0.08)	(0.15)	(0.06)	(0.12)	(0.04)	(0.09)	(0.07)	(0.13)	(0.06)	(0.13)
# news, across	−0.00	−0.10*	−0.01	0.00	0.00	−0.10⁎⁎	0.02	−0.09*	−0.01	0.01	0.01	−0.10⁎⁎
	(0.03)	(0.06)	(0.01)	(0.01)	(0.02)	(0.05)	(0.03)	(0.05)	(0.01)	(0.01)	(0.02)	(0.05)
Close contacts							0.27⁎⁎⁎		0.47⁎⁎⁎		0.11⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.11⁎⁎⁎		1.40⁎⁎⁎		0.47⁎⁎⁎
								(0.05)		(0.04)		(0.07)
N	1705	1705	899	899	806	806	1705	1705	899	899	806	806
Panel C: 7 days lag
# news, within	0.81⁎⁎⁎	0.56⁎⁎⁎	−0.04	0.36*	1.05⁎⁎⁎	0.85⁎⁎⁎	0.66⁎⁎⁎	−0.39⁎⁎⁎	−0.21⁎⁎	0.42⁎⁎	0.95⁎⁎⁎	0.35*
	(0.06)	(0.12)	(0.12)	(0.21)	(0.09)	(0.18)	(0.06)	(0.12)	(0.10)	(0.18)	(0.09)	(0.20)
# news2, within	−0.54⁎⁎⁎	−0.33⁎⁎⁎	0.20⁎⁎	−0.18	−0.75⁎⁎⁎	−0.55⁎⁎⁎	−0.45⁎⁎⁎	0.30⁎⁎⁎	0.28⁎⁎⁎	−0.47⁎⁎⁎	−0.69⁎⁎⁎	−0.19
	(0.04)	(0.09)	(0.10)	(0.17)	(0.06)	(0.13)	(0.04)	(0.09)	(0.08)	(0.15)	(0.06)	(0.14)
# news, across	−0.00	0.02	−0.01	−0.01	0.00	0.02*	−0.01	0.02	0.00	−0.01	−0.00	0.02*
	(0.01)	(0.01)	(0.01)	(0.02)	(0.00)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.00)	(0.01)
Close contacts							0.28⁎⁎⁎		0.47⁎⁎⁎		0.11⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.17⁎⁎⁎		1.46⁎⁎⁎		0.48⁎⁎⁎
								(0.05)		(0.05)		(0.07)
N	1643	1643	837	837	806	806	1643	1643	837	837	806	806
Province FE	√	√	√	√	√	√	√	√	√	√	√	√
Date FE	√	√	√	√	√	√	√	√	√	√	√	√
Controls	√	√	√	√	√	√	√	√	√	√	√	√

Notes: 1. Standard errors in parentheses; 2. * p<0.1

p<0.05

p<0.01; 3. T=the full sample, T=subsample with data from Jan 1 to Feb 5, T=subsample with data from Feb 6 to Feb 29; 4. R=new cases, C=Close contacts, # news: number of news releases and reports.

The effects of population mobility

Tian et al. (2020) confirmed that the number of cases in a province has a strong and positive correlation with the population outflow from Wuhan in the early stage of the pandemic. To test whether the effects of control policies can be mediated by population mobility, we estimate Eqs. (9)-(12) with data of within- and across-province population inflow and movement. The within province population movement may have a limited impact on disease spread in other provinces, and we only included the population inflow into a province. Because population inflow and outflow are often correlated, we used the net inflow to proxy the population movement.

The results for Eqs. (9)-(12) are presented in Table 4 . Consistent with the conventional wisdom, the increase of population movement is associated with a higher number of confirmed cases. An increase of one unit of the population mobility index within a province increases the number of new cases by 14%, and a one-unit change in the inflow mobility index within a province and across provinces will increase the number of new cases by 17% and 1%, respectively. Similarly, a one-unit increase in population mobility index within a province led to a 23% increase in the number of close contacts. The changes in the number of close contacts are 5% and −3% for a 100% increase in population inflow in one province and from other provinces.

Table 4

The effects of within− and across−province population movement on daily spread of COVID−19

	T		T1		T2		T		T1		T2
	R	C	R	C	R	C	R	C	R	C	R	C
Panel A: 3 days lag
inner movement, within	0.14⁎⁎⁎	0.23⁎⁎⁎	0.30⁎⁎⁎	0.21	0.09⁎⁎⁎	0.15⁎⁎	0.08⁎⁎⁎	0.02	0.21⁎⁎⁎	−0.38⁎⁎⁎	0.06⁎⁎	0.10
	(0.03)	(0.06)	(0.08)	(0.17)	(0.03)	(0.06)	(0.02)	(0.06)	(0.06)	(0.13)	(0.03)	(0.06)
population inflow, within	0.17⁎⁎⁎	0.05	0.09	0.24*	0.09⁎⁎⁎	−0.03	0.15⁎⁎⁎	−0.21⁎⁎⁎	−0.02	0.07	0.10⁎⁎⁎	−0.09
	(0.03)	(0.06)	(0.06)	(0.13)	(0.03)	(0.06)	(0.02)	(0.06)	(0.04)	(0.09)	(0.03)	(0.06)
population inflow, across	0.01	−0.03*	0.01⁎⁎⁎	0.01	0.00	−0.03⁎⁎	0.01⁎⁎	−0.04⁎⁎	0.00⁎⁎	−0.01⁎⁎	0.01	−0.03⁎⁎
	(0.01)	(0.02)	(0.00)	(0.01)	(0.01)	(0.02)	(0.01)	(0.02)	(0.00)	(0.00)	(0.01)	(0.01)
Close contacts							0.28⁎⁎⁎		0.44⁎⁎⁎		0.17⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.51⁎⁎⁎		1.95⁎⁎⁎		0.67⁎⁎⁎
								(0.06)		(0.05)		(0.09)
N	1153	1153	677	677	476	476	1153	1153	677	677	476	476
Panel B: 5 days lag
inner movement, within	0.11⁎⁎⁎	0.19⁎⁎⁎	0.07	0.01	0.07⁎⁎⁎	0.14⁎⁎	0.05⁎⁎	0.03	0.06	−0.11	0.05*	0.09
	(0.03)	(0.06)	(0.09)	(0.17)	(0.03)	(0.05)	(0.02)	(0.05)	(0.07)	(0.13)	(0.03)	(0.05)
population inflow, within	0.10⁎⁎⁎	0.03	−0.01	0.13	0.04	0.03	0.09⁎⁎⁎	−0.12⁎⁎	−0.07	0.15	0.04	0.00
	(0.03)	(0.06)	(0.07)	(0.13)	(0.03)	(0.06)	(0.02)	(0.05)	(0.05)	(0.10)	(0.03)	(0.06)
population inflow, across	0.01	−0.04	0.01*	−0.00	0.00	−0.06⁎⁎	0.03⁎⁎	−0.06⁎⁎	0.01⁎⁎	−0.01⁎⁎	0.01	−0.06⁎⁎
	(0.01)	(0.03)	(0.00)	(0.01)	(0.01)	(0.03)	(0.01)	(0.03)	(0.00)	(0.00)	(0.01)	(0.02)
Close contacts							0.32⁎⁎⁎		0.48⁎⁎⁎		0.19⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.47⁎⁎⁎		1.79⁎⁎⁎		0.69⁎⁎⁎
								(0.05)		(0.05)		(0.09)
N	1127	1127	673	673	454	454	1127	1127	673	673	454	454
Panel C: 7 days lag
inner movement, within	0.13⁎⁎⁎	0.28⁎⁎⁎	−0.13	−0.35	0.10⁎⁎⁎	0.19⁎⁎⁎	0.04	0.10*	0.04	−0.14	0.06*	0.12⁎⁎
	(0.03)	(0.06)	(0.12)	(0.22)	(0.03)	(0.06)	(0.03)	(0.06)	(0.10)	(0.17)	(0.03)	(0.06)
population inflow, within	0.07⁎⁎	−0.08	0.03	0.21*	0.02	−0.04	0.09⁎⁎⁎	−0.17⁎⁎⁎	−0.08	0.16*	0.03	−0.05
	(0.03)	(0.06)	(0.07)	(0.12)	(0.04)	(0.06)	(0.03)	(0.06)	(0.05)	(0.10)	(0.03)	(0.06)
population inflow, across	0.02	−0.03	−0.00	−0.00	0.01	−0.03	0.03⁎⁎	−0.05⁎⁎	−0.00	0.00	0.02	−0.04*
	(0.01)	(0.03)	(0.00)	(0.01)	(0.01)	(0.02)	(0.01)	(0.02)	(0.00)	(0.00)	(0.01)	(0.02)
Close contacts							0.32⁎⁎⁎		0.50⁎⁎⁎		0.20⁎⁎⁎
							(0.01)		(0.01)		(0.03)
New cases								1.35⁎⁎⁎		1.59⁎⁎⁎		0.67⁎⁎⁎
								(0.05)		(0.05)		(0.08)
N	1096	1096	664	664	432	432	1096	1096	664	664	432	432
Province FE	√	√	√	√	√	√	√	√	√	√	√	√
Date FE	√	√	√	√	√	√	√	√	√	√	√	√
Controls	√	√	√	√	√	√	√	√	√	√	√	√

Notes: 1. Standard errors in parentheses; 2. * p<0.1

p<0.05

p<0.01; 3. T=the full sample, T=subsample with data from Jan 1 to Feb 5, T=subsample with data from Feb 6 to Feb 29; 4. R=new cases, C=Close contacts.

The mediate effect of media coverage through population mobility

Media coverage may reduce the intensity of population mobility and increase adherence to the mandates of facemask wearing and social distancing. To test this hypothesis that population mobility is a mediating factor for media coverage, we follow the strategy of Baron and Kenny (1986) to regress the within- and across-province population mobility on media coverage. The results are presented in Appendix Table D1 . Media coverage will reduce the intensity of within- and across-province population mobility, and the effects are stronger in the early stage. Within-province population mobility is not only controlled by the within-province media coverage but also affected by media coverage of other provinces that may increase population inflow.

Table D1

The effects of media coverage on within− and across−province population movement

	Population inflow, across provinces			Population movement, within province
	(1)	(2)	(3)	(4)	(5)	(6)
Panel A: within province model
# news	−0.019	−0.043⁎⁎	−0.014	−0.004⁎⁎	−0.016⁎⁎	−0.003
	(0.012)	(0.019)	(0.012)	(0.002)	(0.006)	(0.009)
N	1860	930	930	1860	930	930
Panel B: across provinces model
# news, within	−0.0118	−0.042⁎⁎	−0.013	−0.005⁎⁎⁎	−0.021⁎⁎⁎	−0.002
	(0.012)	(0.018)	(0.011)	(0.002)	(0.006)	(0.009)
# news, across	0.001	0.001	0.001*	−0.001⁎⁎⁎	−0.004⁎⁎⁎	0.001⁎⁎
	(0.004)	(0.008)	(0.005)	(0.001)	(0.002)	(0.003)
N	1860	930	930	1860	930	930
Province FE	√	√	√	√	√	√
Date FE	√	√	√	√	√	√
Controls	√	√	√	√	√	√

Notes: 1. Standard errors in parentheses; 2. * p<0.1

p<0.05

p<0.01; 3. #news: cumulative number of news releases and reports.

We estimate the Eqs. (9) and (12) with media coverage and population mobility. The results are reported in Table 5 , which shows the increasing effects of media coverage and the decreasing effects of population mobility. Several coefficients of population mobility have changed from positive to negative, reflecting that people may move out of population centers that had experienced high incidence rates of COVID-19.

Table 5

The direct and mediating effects of media coverage on daily spread of COVID−19 across provinces

	T		T1		T2		T		T1		T2
	R	C	R	C	R	C	R	C	R	C	R	C
Panel A: 3 days lag
# news, within	1.18⁎⁎⁎	1.32⁎⁎⁎	−1.67⁎⁎⁎	−0.59	1.27⁎⁎⁎	1.27⁎⁎⁎	0.84⁎⁎⁎	−0.40	−1.41⁎⁎⁎	2.69⁎⁎⁎	1.13⁎⁎⁎	0.66⁎⁎
	(0.10)	(0.25)	(0.42)	(0.88)	(0.14)	(0.29)	(0.10)	(0.25)	(0.30)	(0.64)	(0.14)	(0.31)
# news2, within	−0.77⁎⁎⁎	−0.80⁎⁎⁎	1.23⁎⁎⁎	0.73	−0.85⁎⁎⁎	−0.79⁎⁎⁎	−0.57⁎⁎⁎	0.33⁎⁎	0.91⁎⁎⁎	−1.69⁎⁎⁎	−0.76⁎⁎⁎	−0.38*
	(0.07)	(0.17)	(0.29)	(0.62)	(0.09)	(0.19)	(0.07)	(0.17)	(0.21)	(0.45)	(0.09)	(0.20)
# news, across	0.00	−0.00	0.01	0.09*	0.01	−0.01	0.01	−0.01	−0.03	0.07*	0.01	−0.01
	(0.01)	(0.01)	(0.02)	(0.05)	(0.01)	(0.01)	(0.01)	(0.01)	(0.02)	(0.04)	(0.01)	(0.01)
inner movement, within	0.02	0.02	0.38⁎⁎⁎	−0.17	−0.00	0.01	0.02	−0.01	0.46⁎⁎⁎	−0.92⁎⁎⁎	−0.00	0.01
	(0.03)	(0.07)	(0.11)	(0.23)	(0.03)	(0.06)	(0.03)	(0.06)	(0.08)	(0.17)	(0.03)	(0.06)
population inflow, within	−0.00	−0.21⁎⁎⁎	0.32⁎⁎⁎	0.11	−0.03	−0.21⁎⁎⁎	0.05*	−0.21⁎⁎⁎	0.27⁎⁎⁎	−0.51⁎⁎⁎	−0.01	−0.19⁎⁎⁎
	(0.03)	(0.07)	(0.10)	(0.20)	(0.04)	(0.07)	(0.03)	(0.07)	(0.07)	(0.15)	(0.04)	(0.07)
population inflow, across	−0.00	−0.02	0.01⁎⁎⁎	0.01⁎⁎	−0.01	−0.02	0.00	−0.02	0.00	−0.00	−0.00	−0.01
	(0.01)	(0.02)	(0.00)	(0.01)	(0.01)	(0.02)	(0.01)	(0.02)	(0.00)	(0.00)	(0.01)	(0.02)
Close contacts							0.25⁎⁎⁎		0.44⁎⁎⁎		0.11⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.46⁎⁎⁎		1.96⁎⁎⁎		0.48⁎⁎⁎
								(0.06)		(0.05)		(0.09)
N	1153	1153	677	677	476	476	1153	1153	677	677	476	476
Panel B: 5 days lag
# news, within	0.77⁎⁎⁎	1.19⁎⁎⁎	−0.16	−0.26	0.98⁎⁎⁎	1.39⁎⁎⁎	0.41⁎⁎⁎	0.11	−0.04	0.02	0.78⁎⁎⁎	0.84⁎⁎⁎
	(0.11)	(0.24)	(0.42)	(0.81)	(0.15)	(0.29)	(0.11)	(0.23)	(0.32)	(0.61)	(0.15)	(0.30)
# news2, within	−0.49⁎⁎⁎	−0.71⁎⁎⁎	0.29	0.58	−0.67⁎⁎⁎	−0.87⁎⁎⁎	−0.27⁎⁎⁎	−0.02	0.02	0.07	−0.54⁎⁎⁎	−0.50⁎⁎
	(0.08)	(0.16)	(0.30)	(0.59)	(0.10)	(0.20)	(0.07)	(0.15)	(0.23)	(0.45)	(0.10)	(0.20)
# news, across	0.01	−0.02	−0.05⁎⁎⁎	−0.07⁎⁎⁎	0.01	−0.02	0.01*	−0.03*	−0.02⁎⁎	0.02	0.01	−0.02
	(0.01)	(0.02)	(0.01)	(0.02)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.02)	(0.01)	(0.01)
inner movement, within	0.02	−0.02	−0.09	−0.33*	0.02	−0.01	0.03	−0.05	0.06	−0.16	0.02	−0.02
	(0.03)	(0.06)	(0.10)	(0.19)	(0.03)	(0.06)	(0.03)	(0.06)	(0.08)	(0.15)	(0.03)	(0.06)
population inflow, within	−0.02	−0.22⁎⁎⁎	−0.22*	−0.29	−0.04	−0.14⁎⁎	0.05*	−0.19⁎⁎⁎	−0.08	0.09	−0.02	−0.12*
	(0.03)	(0.07)	(0.11)	(0.22)	(0.03)	(0.07)	(0.03)	(0.06)	(0.09)	(0.17)	(0.03)	(0.06)
population inflow, across	0.01	−0.03	−0.00	−0.01	−0.00	−0.05⁎⁎	0.02*	−0.04*	0.00	−0.00	0.01	−0.04⁎⁎
	(0.01)	(0.02)	(0.00)	(0.00)	(0.01)	(0.02)	(0.01)	(0.02)	(0.00)	(0.00)	(0.01)	(0.02)
Close contacts							0.30⁎⁎⁎		0.47⁎⁎⁎		0.15⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.40⁎⁎⁎		1.77⁎⁎⁎		0.55⁎⁎⁎
								(0.06)		(0.05)		(0.09)
N	1127	1127	673	673	454	454	1127	1127	673	673	454	454
Panel C: 7 days lag
# news, within	0.84⁎⁎⁎	0.88⁎⁎⁎	−0.83*	−1.17	1.17⁎⁎⁎	1.03⁎⁎⁎	0.57⁎⁎⁎	−0.25	−0.24	0.12	1.01⁎⁎⁎	0.32
	(0.12)	(0.26)	(0.48)	(0.86)	(0.17)	(0.32)	(0.12)	(0.24)	(0.39)	(0.69)	(0.17)	(0.34)
# news2, within	−0.55⁎⁎⁎	−0.55⁎⁎⁎	0.84⁎⁎	1.43⁎⁎	−0.82⁎⁎⁎	−0.69⁎⁎⁎	−0.38⁎⁎⁎	0.19	0.13	0.12	−0.70⁎⁎⁎	−0.20
	(0.08)	(0.17)	(0.40)	(0.70)	(0.11)	(0.22)	(0.08)	(0.16)	(0.32)	(0.56)	(0.11)	(0.23)
# news, across	0.03	−0.03	0.01⁎⁎⁎	0.02⁎⁎⁎	0.01	−0.03	0.04*	−0.07*	0.00	0.00⁎⁎	0.02	−0.04
	(0.02)	(0.04)	(0.00)	(0.00)	(0.02)	(0.04)	(0.02)	(0.04)	(0.00)	(0.00)	(0.02)	(0.04)
inner movement, within	0.06*	0.12*	−0.31⁎⁎	−0.81⁎⁎⁎	0.05	0.10	0.02	0.04	0.10	−0.33*	0.04	0.06
	(0.03)	(0.07)	(0.13)	(0.24)	(0.03)	(0.07)	(0.03)	(0.07)	(0.11)	(0.19)	(0.03)	(0.07)
population inflow, within	0.02	−0.31⁎⁎	−0.01	−0.05	−0.04	−0.21*	0.11⁎⁎	−0.33⁎⁎⁎	0.02	−0.03	−0.00	−0.19*
	(0.06)	(0.13)	(0.11)	(0.19)	(0.06)	(0.11)	(0.06)	(0.12)	(0.09)	(0.15)	(0.06)	(0.11)
population inflow, across	0.07	−0.09	−0.01⁎⁎⁎	−0.02⁎⁎⁎	0.03	−0.09	0.10⁎⁎	−0.18*	−0.00	−0.00	0.04	−0.11
	(0.05)	(0.11)	(0.00)	(0.01)	(0.05)	(0.09)	(0.05)	(0.10)	(0.00)	(0.01)	(0.05)	(0.09)
Close contacts							0.31⁎⁎⁎		0.50⁎⁎⁎		0.16⁎⁎⁎
							(0.01)		(0.02)		(0.02)
New cases								1.34⁎⁎⁎		1.56⁎⁎⁎		0.60⁎⁎⁎
								(0.05)		(0.05)		(0.09)
N	1096	1096	664	664	432	432	1096	1096	664	664	432	432
Province FE	√	√	√	√	√	√	√	√	√	√	√	√
Date FE	√	√	√	√	√	√	√	√	√	√	√	√
Controls	√	√	√	√	√	√	√	√	√	√	√	√

Notes: 1. Standard errors in parentheses; 2. * p<0.1

p<0.05

p<0.01; 3. T=the full sample, T=subsample with data from Jan 1 to Feb 5, T=subsample with data from Feb 6 to Feb 29; 4. R=new cases, C=Close contacts, # news: number of news releases and reports.

Robustness Check with Hubei excluded

As a robustness check, we estimated the models with the sample excluding the province of Hubei. The results of the set of regressions on the sample excluding Hubei are in Appendix D (Appendix Table D2, Table D3, Table D4, Table D5 ). For the baseline model, while the main results remain the same, it appears that the epidemic transmission is slightly weaker in the provinces other than Hubei except for spatial transmission. That may be because of delayed and less intensive media coverage in provinces other than Hubei. For the model on media coverage, the sample with Hubei shows stronger impacts of media coverage, potentially because the other provinces saw the situation in Wuhan and were more informed and organized than Wuhan at the initial stage of the epidemic.

Table D2

Daily spread of COVID-19 within province excluding Hubei

	T		T1		T2		T		T1		T2
	R	C	R	C	R	C	R	C	R	C	R	C
Panel A: 3 days lag
ɑwithin,	0.19⁎⁎⁎	0.24⁎⁎⁎	0.28⁎⁎⁎	0.34⁎⁎⁎	0.10⁎⁎⁎	0.16⁎⁎⁎	0.17⁎⁎⁎	0.01	0.15⁎⁎⁎	−0.08⁎⁎⁎	0.08⁎⁎⁎	0.12⁎⁎⁎
	(0.01)	(0.02)	(0.01)	(0.03)	(0.02)	(0.03)	(0.01)	(0.02)	(0.01)	(0.03)	(0.02)	(0.03)
Close contacts							0.26⁎⁎⁎		0.44⁎⁎⁎		0.12⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.09⁎⁎⁎		1.37⁎⁎⁎		0.47⁎⁎⁎
								(0.05)		(0.04)		(0.07)
N	1710	1710	930	930	780	780	1767	1767	961	961	806	806
Panel B: 5 days lag
ɑwithin	0.15⁎⁎⁎	0.24⁎⁎⁎	0.24⁎⁎⁎	0.31⁎⁎⁎	0.06⁎⁎⁎	0.22⁎⁎⁎	0.13⁎⁎⁎	0.05*	0.11⁎⁎⁎	−0.06⁎⁎	0.03⁎⁎	0.20⁎⁎⁎
	(0.01)	(0.02)	(0.02)	(0.03)	(0.02)	(0.03)	(0.01)	(0.02)	(0.02)	(0.03)	(0.02)	(0.03)
Close contacts							0.27⁎⁎⁎		0.47⁎⁎⁎		0.13⁎⁎⁎
							(0.01)		(0.02)		(0.02)
New cases								1.08⁎⁎⁎		1.37⁎⁎⁎		0.49⁎⁎⁎
								(0.04)		(0.04)		(0.07)
N	1650	1650	870	870	780	780	1705	1705	899	899	806	806
Panel C: 7 days lag
ɑwithin	0.12⁎⁎⁎	0.15⁎⁎⁎	0.26⁎⁎⁎	0.25⁎⁎⁎	0.12⁎⁎⁎	0.11⁎⁎⁎	0.11⁎⁎⁎	−0.01	0.15⁎⁎⁎	−0.14⁎⁎⁎	0.01	0.10⁎⁎⁎
	(0.02)	(0.03)	(0.04)	(0.04)	(0.02)	(0.04)	(0.01)	(0.02)	(0.02)	(0.03)	(0.02)	(0.04)
Close contacts							0.28⁎⁎⁎		0.47⁎⁎⁎		0.15⁎⁎⁎
							(0.01)	1.14⁎⁎⁎	(0.01)	1.43⁎⁎⁎	(0.02)	0.57⁎⁎⁎
New cases								(0.04)		(0.05)		(0.07)
N	1590	1590	810	810	780	780	1643	1643	837	837	806	806
Province FE	√	√	√	√	√	√	√	√	√	√	√	√
Date FE	√	√	√	√	√	√	√	√	√	√	√	√
Controls	×	×	×	×	×	×	×	×	×	×	˟	˟

Notes: 1. Standard errors in parentheses; 2. * p<0.1

p<0.05

p<0.01; 3. T=the full sample, T=subsample with data from Jan 1 to Feb 5, T=subsample with data from Feb 6 to Feb 29; 4. R=New cases, C=Close contacts.

Table D3

Spread of COVID-19 across provinces, excluding Hubei

	T		T1		T2		T		T1		T2
	R	C	R	C	R	C	R	C	R	C	R	C
Panel A: 3 days lag
ɑwithin	0.18⁎⁎⁎	0.19⁎⁎⁎	0.27⁎⁎⁎	0.34⁎⁎⁎	0.10⁎⁎⁎	0.12⁎⁎⁎	0.14⁎⁎⁎	−0.01	0.13⁎⁎⁎	−0.07*	0.08⁎⁎⁎	0.07⁎⁎
	(0.01)	(0.03)	(0.02)	(0.04)	(0.02)	(0.04)	(0.01)	(0.03)	(0.02)	(0.04)	(0.02)	(0.04)
ɑacros	−0.01	−0.04⁎⁎	−0.01	0.00	−0.01	−0.04⁎⁎	0.00	−0.03*	−0.01	0.02	−0.00	−0.04⁎⁎
	(0.01)	(0.02)	(0.02)	(0.03)	(0.01)	(0.02)	(0.01)	(0.02)	(0.01)	(0.03)	(0.01)	(0.02)
Close contacts							0.21⁎⁎⁎		0.40⁎⁎⁎		0.12⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.13⁎⁎⁎		1.54⁎⁎⁎		0.50⁎⁎⁎
								(0.05)		(0.05)		(0.07)
N	1710	1710	930	930	780	780	1710	1710	930	930	780	780
Panel B: 5 days lag
ɑwithin	0.16⁎⁎⁎	0.25⁎⁎⁎	0.24⁎⁎⁎	0.32⁎⁎⁎	0.04	−0.17	0.11⁎⁎⁎	0.07⁎⁎⁎	0.10⁎⁎⁎	−0.04	0.07	−0.20
	(0.01)	(0.02)	(0.02)	(0.04)	(0.10)	(0.19)	(0.01)	(0.02)	(0.02)	(0.04)	(0.10)	(0.19)
ɑacros	0.01⁎⁎⁎	0.02⁎⁎⁎	−0.01	0.00	−0.02	−0.39⁎⁎	0.01⁎⁎⁎	0.00*	−0.01	0.02	0.03	−0.38⁎⁎
	(0.00)	(0.00)	(0.01)	(0.03)	(0.10)	(0.18)	(0.00)	(0.00)	(0.01)	(0.02)	(0.09)	(0.18)
Close contacts							0.22⁎⁎⁎		0.42⁎⁎⁎		0.14⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.11⁎⁎⁎		1.52⁎⁎⁎		0.52⁎⁎⁎
								(0.05)		(0.05)		(0.07)
N	1650	1650	870	870	780	780	1650	1650	870	870	780	780
Panel C: 3 days lag
ɑwithin	0.12⁎⁎⁎	0.19⁎⁎⁎	0.25⁎⁎⁎	0.23⁎⁎⁎	0.03	0.14⁎⁎⁎	0.08⁎⁎⁎	0.04	0.15⁎⁎⁎	−0.16⁎⁎⁎	0.01	0.12⁎⁎⁎
	(0.01)	(0.03)	(0.02)	(0.04)	(0.02)	(0.04)	(0.01)	(0.03)	(0.02)	(0.04)	(0.02)	(0.04)
ɑacros	−0.00	0.03	−0.01	−0.02	0.00	0.03*	−0.01	0.03*	−0.00	−0.00	−0.00	0.03*
	(0.01)	(0.02)	(0.01)	(0.02)	(0.01)	(0.02)	(0.01)	(0.02)	(0.01)	(0.01)	(0.01)	(0.02)
Close contacts							0.23⁎⁎⁎		0.42⁎⁎⁎		0.15⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.17⁎⁎⁎		1.59⁎⁎⁎		0.60⁎⁎⁎
								(0.05)		(0.05)		(0.07)
N	1590	1590	810	810	780	780	1590	1590	810	810	780	780
Province FE	√	√	√	√	√	√	√	√	√	√	√	√
Date FE	√	√	√	√	√	√	√	√	√	√	√	√
Controls	×	×	×	×	×	×	×	×	×	×	˟	˟

Notes: 1. Standard errors in parentheses; 2. * p<0.1

p<0.05

p<0.01; 3. T=the full sample, T=subsample with data from Jan 1 to Feb 5, T=subsample with data from Feb 6 to Feb 29; 4. R=New cases, C=Close contacts.

Table D4

The effects of media coverage on spread of COVID-19 across provinces, excluding Hubei

	T		T1		T2		T		T1		T2
	R	C	R	C	R	C	R	C	R	C	R	C
Panel A: 3 days lag
# news, within	1.03⁎⁎⁎	0.81⁎⁎⁎	0.69⁎⁎⁎	0.83⁎⁎⁎	1.24⁎⁎⁎	0.78⁎⁎⁎	0.88⁎⁎⁎	−0.38⁎⁎⁎	0.36⁎⁎⁎	−0.24	1.17⁎⁎⁎	0.27
	(0.05)	(0.12)	(0.09)	(0.17)	(0.08)	(0.18)	(0.05)	(0.13)	(0.08)	(0.16)	(0.08)	(0.20)
# news2, within	−0.69⁎⁎⁎	−0.51⁎⁎⁎	−0.39⁎⁎⁎	−0.46⁎⁎⁎	−0.85⁎⁎⁎	−0.51⁎⁎⁎	−0.59⁎⁎⁎	0.28⁎⁎⁎	−0.21⁎⁎⁎	0.15	−0.80⁎⁎⁎	−0.16
	(0.03)	(0.09)	(0.07)	(0.14)	(0.06)	(0.12)	(0.03)	(0.09)	(0.06)	(0.13)	(0.06)	(0.14)
# news, across	−0.00	−0.02⁎⁎	−0.01	0.00	−0.00	−0.02⁎⁎	0.00	−0.02*	−0.01	0.01	0.00	−0.02⁎⁎
	(0.00)	(0.01)	(0.01)	(0.02)	(0.00)	(0.01)	(0.00)	(0.01)	(0.01)	(0.02)	(0.00)	(0.01)
Close contacts							0.19⁎⁎⁎		0.40⁎⁎⁎		0.09⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.15⁎⁎⁎		1.55⁎⁎⁎		0.41⁎⁎⁎
								(0.06)		(0.05)		(0.08)
N	1710	1710	930	930	780	780	1710	1710	930	930	780	780
Panel B: 5 days lag
# news, within	0.88⁎⁎⁎	0.61⁎⁎⁎	0.55⁎⁎⁎	0.87⁎⁎⁎	1.04⁎⁎⁎	0.67⁎⁎⁎	0.75⁎⁎⁎	−0.40⁎⁎⁎	0.18⁎⁎	0.04	0.96⁎⁎⁎	0.15
	(0.06)	(0.14)	(0.10)	(0.19)	(0.09)	(0.19)	(0.06)	(0.14)	(0.09)	(0.17)	(0.09)	(0.20)
# news2, within	−0.59⁎⁎⁎	−0.41⁎⁎⁎	−0.30⁎⁎⁎	−0.52⁎⁎⁎	−0.72⁎⁎⁎	−0.46⁎⁎⁎	−0.51⁎⁎⁎	0.26⁎⁎⁎	−0.08	−0.07	−0.67⁎⁎⁎	−0.09
	(0.04)	(0.09)	(0.08)	(0.16)	(0.06)	(0.13)	(0.04)	(0.09)	(0.07)	(0.14)	(0.06)	(0.14)
# news, across	0.00	−0.10*	−0.00	0.00	0.00	−0.10⁎⁎	0.02	−0.10*	−0.01	0.01	0.01	−0.10⁎⁎
	(0.02)	(0.06)	(0.01)	(0.01)	(0.02)	(0.05)	(0.02)	(0.05)	(0.01)	(0.01)	(0.02)	(0.05)
Close contacts							0.21⁎⁎⁎		0.42⁎⁎⁎		0.12⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.14⁎⁎⁎		1.52⁎⁎⁎		0.50⁎⁎⁎
								(0.05)		(0.05)		(0.07)
N	1650	1650	870	870	780	780	1650	1650	870	870	780	780
Panel C: 7 days lag
# news, within	0.86⁎⁎⁎	0.60⁎⁎⁎	0.51⁎⁎⁎	0.81⁎⁎⁎	1.07⁎⁎⁎	0.79⁎⁎⁎	0.73⁎⁎⁎	−0.44⁎⁎⁎	0.17	−0.01	0.98⁎⁎⁎	0.23
	(0.06)	(0.13)	(0.12)	(0.23)	(0.09)	(0.19)	(0.05)	(0.13)	(0.10)	(0.20)	(0.09)	(0.21)
# news2, within	−0.60⁎⁎⁎	−0.37⁎⁎⁎	−0.27⁎⁎⁎	−0.54⁎⁎⁎	−0.77⁎⁎⁎	−0.51⁎⁎⁎	−0.52⁎⁎⁎	0.35⁎⁎⁎	−0.04	−0.11	−0.71⁎⁎⁎	−0.11
	(0.04)	(0.10)	(0.10)	(0.20)	(0.06)	(0.13)	(0.04)	(0.10)	(0.09)	(0.17)	(0.06)	(0.14)
# news, across	0.00	0.02	−0.01	−0.02	0.00	0.02*	−0.00	0.02	−0.00	−0.00	−0.00	0.02*
	(0.01)	0.60⁎⁎⁎	(0.01)	0.81⁎⁎⁎	(0.01)	0.79⁎⁎⁎	(0.00)	(0.01)	(0.01)	(0.01)	(0.00)	(0.01)
Close contacts							0.22⁎⁎⁎		0.43⁎⁎⁎		0.11⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.21⁎⁎⁎		1.58⁎⁎⁎		0.52⁎⁎⁎
								(0.05)		(0.05)		(0.08)
N	1590	1590	810	810	780	780	1590	1590	810	810	780	780
Province FE	√	√	√	√	√	√	√	√	√	√	√	√
Date FE	√	√	√	√	√	√	√	√	√	√	√	√
Controls	√	√	√	√	√	√	√	√	√	√	√	√

Notes: 1. Standard errors in parentheses; 2. * p<0.1

p<0.05

p<0.01; 3. T=the full sample, T=subsample with data from Jan 1 to Feb 5, T=subsample with data from Feb 6 to Feb 29; 4. R=New cases, C=Close contacts, #news: cumulative number of news releases and reports.

Table D5

The effects of within- and across-provinces population mobility on spread of COVID-19, excluding Hubei

	T		T1		T2		T		T1		T2
	R	C	R	C	R	C	R	C	R	C	R	C
Panel A: 3 days lag
inner movement, within	0.13⁎⁎⁎	0.22⁎⁎⁎	0.17⁎⁎	0.22	0.09⁎⁎⁎	0.15⁎⁎	0.08⁎⁎⁎	0.00	0.08	−0.15	0.07⁎⁎	0.09
	(0.02)	(0.06)	(0.08)	(0.17)	(0.03)	(0.06)	(0.02)	(0.06)	(0.05)	(0.11)	(0.03)	(0.06)
population inflow, within	0.09⁎⁎⁎	−0.02	0.06	0.24⁎⁎	0.09⁎⁎	−0.03	0.09⁎⁎⁎	−0.16⁎⁎⁎	−0.04	0.10	0.09⁎⁎⁎	−0.09
	(0.02)	(0.06)	(0.05)	(0.12)	(0.03)	(0.07)	(0.02)	(0.06)	(0.03)	(0.08)	(0.03)	(0.07)
population inflow, across	0.00	−0.03⁎⁎	0.01⁎⁎⁎	0.01	0.00	−0.03⁎⁎	0.01	−0.03⁎⁎	0.00⁎⁎⁎	−0.01⁎⁎⁎	0.01	−0.03⁎⁎
	(0.01)	(0.02)	(0.00)	(0.01)	(0.01)	(0.02)	(0.01)	(0.02)	(0.00)	(0.00)	(0.01)	(0.02)
Close contacts							0.24⁎⁎⁎		0.43⁎⁎⁎		0.18⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.66⁎⁎⁎		2.18⁎⁎⁎		0.68⁎⁎⁎
								(0.07)		(0.04)		(0.09)
N	1132	1132	663	663	469	469	1132	1132	663	663	469	469
Panel B: 5 days lag
inner movement, within	0.11⁎⁎⁎	0.19⁎⁎⁎	0.46⁎⁎⁎	0.57⁎⁎	0.08⁎⁎⁎	0.13⁎⁎	0.06⁎⁎⁎	0.02	0.21⁎⁎⁎	−0.36⁎⁎	0.05*	0.08
	(0.02)	(0.06)	(0.10)	(0.22)	(0.03)	(0.06)	(0.02)	(0.05)	(0.07)	(0.16)	(0.03)	(0.06)
population inflow, within	0.05⁎⁎	−0.00	0.02	0.21	0.04	0.03	0.05⁎⁎	−0.08	−0.07	0.16	0.03	0.00
	(0.02)	(0.06)	(0.06)	(0.14)	(0.03)	(0.06)	(0.02)	(0.06)	(0.05)	(0.10)	(0.03)	(0.06)
population inflow, across	0.00	−0.05*	0.01⁎⁎	0.00	0.00	−0.06⁎⁎	0.02	−0.06⁎⁎	0.01⁎⁎⁎	−0.01⁎⁎⁎	0.01	−0.06⁎⁎
	(0.01)	(0.03)	(0.00)	(0.01)	(0.01)	(0.03)	(0.01)	(0.03)	(0.00)	(0.00)	(0.01)	(0.02)
Close contacts							0.26⁎⁎⁎		0.43⁎⁎⁎		0.19⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.59⁎⁎⁎		2.03⁎⁎⁎		0.71⁎⁎⁎
								(0.07)		(0.05)		(0.09)
N	1105	1105	658	658	447	447	1105	1105	658	658	447	447
Panel C: 7 days lag
inner movement, within	0.12⁎⁎⁎	0.27⁎⁎⁎	0.18	0.11	0.11⁎⁎⁎	0.19⁎⁎⁎	0.04*	0.10*	0.13	−0.20	0.07⁎⁎	0.11*
	(0.03)	(0.06)	(0.13)	(0.24)	(0.03)	(0.06)	(0.03)	(0.06)	(0.10)	(0.19)	(0.03)	(0.06)
population inflow, within	0.03	−0.11*	0.19⁎⁎⁎	0.47⁎⁎⁎	0.02	−0.04	0.06⁎⁎	−0.15⁎⁎⁎	−0.02	0.13	0.02	−0.05
	(0.03)	(0.06)	(0.07)	(0.14)	(0.04)	(0.07)	(0.03)	(0.06)	(0.06)	(0.11)	(0.04)	(0.06)
population inflow, across	0.01	−0.04	−0.01⁎⁎⁎	−0.01	0.01	−0.03	0.02*	−0.05⁎⁎	−0.00*	0.01	0.02	−0.04*
	(0.01)	(0.03)	(0.00)	(0.01)	(0.01)	(0.02)	(0.01)	(0.02)	(0.00)	(0.01)	(0.01)	(0.02)
Close contacts							0.28⁎⁎⁎		0.47⁎⁎⁎		0.20⁎⁎⁎
							(0.01)		(0.01)		(0.03)
New cases								1.42⁎⁎⁎		1.73⁎⁎⁎		0.68⁎⁎⁎
								(0.06)		(0.05)		(0.09)
N	1075	1075	650	650	425	425	1075	1075	650	650	425	425
Province FE	√	√	√	√	√	√	√	√	√	√	√	√
Date FE	√	√	√	√	√	√	√	√	√	√	√	√
Controls	√	√	√	√	√	√	√	√	√	√	√	√

Notes: 1. Standard errors in parentheses; 2. * p<0.1

p<0.05

p<0.01; 3. T=the full sample, T=subsample with data from Jan 1 to Feb 5, T=subsample with data from Feb 6 to Feb 29; 4. R=New cases, C=Close contacts.

Does media coverage work?

The impact of media coverage on COVID-19 transmission in China could be assessed by simulating the possible outcomes if media coverages were absent. Qiu et al. (2020) simulated the counterfactual impact of control policies and concluded the potential cases averted was about 1.4 million by February 29, 2020. We use the same counterfactual strategy to simulate what would be if media coverage had been absent.

We follow the method used by Tian et al. (2020) to replace the within- and across-province population mobility index since the launch of the Level I response with the value of the index on the same day and month in 2019. The across population inflow and the within-province population movement have a similar trend of variation in 2019 and 2020 before the activation of the Level I Response but varied much afterward (Appendix Figure B2 and B3). Our counterfactual simulations are based on the discussions in the methodology section, and we limit the dates to from January 19 to February 29. We simulate the counterfactual outcomes where there was no media coverage with the population mobility indices kept as those in 2019 and other control variables unchanged.

Figure B3

Plots of the variation for within population movement for each province in 2019 and 2020

The results of the daily patients in Figure 3 indicate that the counterfactual total cumulative cases during January 19 to February 29 would be 394,032 (95% CI, 354,646 − 434,147), 237,836.3 (95% CI, 221,870 − 253,803), and 181,953 (95% CI, 169,095 − 194,811) for the 3, 5, and 7-days lag models. It is about 4.9, 3.0, and 2.3 times of the actual number of cases (80,084). Figure 4 reports the results of the counterfactual cumulative number of close contacts as 817,943 (95% CI, 796,157 − 839,730) for the 3-days lag model, 1,082,440 (95% CI, 1,045,493 − 1,119,388) for the 5-days lag model, and 1,434,441 (95% CI, 1,376,103 − 1,492,778) for the 7-day lag model.

Figure 3

Counterfactual Simulations of Absenting News Reports: COVID-19 New Cases

Figure 4

Counterfactual Simulations of Absenting News Reports: Number of Close Contacts

Discussions

This paper uses an augmented SIR model to estimate the COVID-19 transmission in China from January 19 – February 29 and assess the impact of media coverage on the spread of the epidemic after controlling provincial confounding factors and population mobility. Key findings include the following. First, a higher transmission rate during the early stage (between January 1 and February 5) versus the late stage (from February 6 to February 29) is observed. The early-stage of the pandemic saw a stronger effect of the number of close contacts on the number of new cases, potentially resulting from more widespread testing. The number of close contacts associated with additional new cases was higher in the early stage than after February 5, reflecting more stringent prevention and control policies that may have reduced the number of close contacts. Second, the effect of media coverage on the spread of COVID-19 has an inverse-U shape and has a net effect in reducing the number of new cases and close contacts. Third, the increase of within- and across-province population mobility is associated with higher risks of being infected. However, the population mobility may be reduced by increased media coverage of the COVID-19 pandemic. Our counterfactual simulation indicates that media coverage has substantially mitigated the temporal and spatial spread of COVID-19.

Our use of the number of close contacts is new to the literature, and the results have important policy implications. An increase of 100% in the number of close contacts were associated with an increase of 26% in the number of COVID-19 cases during the study period. However, the earlier time period (January 19 - February 5) saw a much stronger correlation, with the associated increase in the number of cases at 44%. In contrast, the percentage was lowered to 12% after February 5, potentially due to increased accessibility of COVID-19 tests and reduced social activities. Similarly, the percentage increase in close contacts due to the number of new cases was 109% for the full time period, 137% before February 5, and 47% afterward. Those results indicate the importance of contact tracing as new cases can be identified and quarantined preeminently. The pattern also provides evidence of China's policies during February 2020 as the two variables (the number of new cases and the number of close contacts) have decoupled. Those policies may include increased testing, social distancing, and the mandate of facemasks wearing. However, we cannot assess the effects of the different components of the prevention and control policies.

This paper has additional implications for the understanding of COVID-19 prevention and control. First, this paper is one of the few works to evaluate the impact of media coverage on COVID-19 transmission. Although prior studies, e.g., Fang et al. (2020), Qiu et al. (2020), and Tian et al. (2020), have estimated the impacts of COVID-19 prevention and control policies in China, the effect of media coverage remains unknown. Second, we provide alternative indicators for the spread of COVID-19. Most available studies used the daily or cumulated number of confirmed cases to measure the spread of COVID-19 (Jia et al., 2020; Qiu et al., 2020; Tian et al., 2020), which describes the variation of disease transmission but cannot portray the spatial dynamics across provinces. Third, we offer additional evidence on the incubation period of COVID-19. Previous studies conclude that the incubation of COVID-19 is about 5.2 days or longer (Guan et al., 2020; Li et al., 2020). We set 3-, 5-, and 7-days incubation in our estimation and find that the transmission would vary across these settings. Our results are confirmed by Zhang et al. (2020), which identified a threshold from a slow- to a fast-growing phase for COVID-19 at 5.5 (95% CI, 4.6–6.4) days after reporting of the symptoms. Fourth, we shed light on the relation between media coverage, population mobility, and COVID-19 transmission. Mobility may be correlated with higher risks of infectious disease transmission (Balcan et al., 2009; Brockmann and Helbing, 2013). Several earlier studies have investigated how the population mobility, which was amplified by the Lunar New Year Holiday, has affected the scale and range of the COVID-19 outbreak (Fang et al., 2020; Jia et al., 2020; Qiu et al., 2020; Tian et al., 2020). We included a novel pathway of the impact of media coverage, i.e., through reduced human mobility. We documented a mediating effect of population mobility where the media coverage reduces within- and across-province population mobility.

This study has two important limitations. First, our measure of media coverage does not measure the extent of the news releases and reports reaching the local population and whether the population in a province would respond to news reports on cases in neighboring provinces. However, as interprovince mobility has dramatically reduced during our study period, the cross-province of media coverage may be limited. Second, we were not able to calculate province-specific impacts of cross-province mobility. We did not differentiate neighboring provinces and non-contiguous provinces – although such differences may be diminished with the wide use of high-speed railway networks and extensive air travels in China.

Conclusions

This paper estimates the transmission of COVID-19 during the early phase of the pandemic in China and the effects of media coverage on the control of the COVID-19 pandemic. Our analysis highlights the importance of contact tracing in containing the COVID-19 pandemic. We have considered within- and across-province transmission and explore whether media coverage was effective, how they work, and the counterfactual impact of absent media coverage. We use the cumulative daily number of official news releases and reports about COVID-19 to measure the media coverage and examine how it is related to the numbers of confirmed cases and close contacts. Our counterfactual simulations suggest that media coverage of COVID-19 in China may have averted 394,000 additional new infections from January 19 to February 29. Our results, along with earlier studies (Chen et al., 2020b; Gharehgozli et al., 2020; Kumar and Managi, 2020; Mandel and Veetil, 2020; Yoo and Managi, 2020), call for concerted efforts and actions to control and mitigate the impact of COVID-19.

Future research may explore the causal pathways between media coverage and reduced COVID-19 transmission, including reduced population mobility, increased adherence to COVID-19 prevention and control measures, e.g., social distancing and wearing of facemasks.

Funding

Dr. Ning Liu and Prof. Guoxian Bao acknowledge support from the National Social Science Foundation of China [Grant No. 20AZD032].

CRediT authorship contribution statement

Ning Liu: Conceptualization, Data curation, Formal analysis, Writing - original draft. Zhuo Chen: Conceptualization, Formal analysis, Methodology, Supervision, Writing - review & editing. Guoxian Bao: Conceptualization, Funding acquisition.

Table A1

Summary Statistics: the number of news reports and releases

Province	Daily number of news reports						Cumulative number of news reports
	Obs	Total Counts	Mean	SD	Min	Max	Obs	Mean	SD	Min	Max
Anhui	36	50	1.39	0.77	1	5	42	31.95	14.79	2	50
Beijing	44	122	2.77	2.73	1	13	45	83.80	34.71	2	122
Chongqing	42	78	1.86	1.20	1	6	43	40.95	25.52	1	78
Fujian	38	58	1.53	0.98	1	5	42	36.45	16.97	1	58
Gansu	19	22	1.16	0.37	1	2	40	14.35	5.55	2	22
Guangdong	37	90	2.43	1.71	1	8	44	58.45	28.77	2	90
Guangxi	37	46	1.24	0.60	1	3	42	26.62	12.93	1	46
Guizhou	37	48	1.30	0.52	1	3	43	27.30	15.10	1	48
Hainan	21	33	1.57	0.68	1	3	27	19.81	10.39	1	33
Hebei	31	36	1.16	0.37	1	2	42	20.76	10.87	1	36
Heilongjiang	37	56	1.51	0.73	1	3	43	33.12	17.07	1	56
Henan	35	45	1.29	0.52	1	3	43	26.37	13.41	1	45
Hubei	54	247	4.57	4.24	1	17	65	119.09	95.39	1	247
Hunan	33	44	1.33	0.65	1	3	42	27.45	13.13	2	44
Inner Mongolia	34	46	1.35	0.69	1	3	41	27.63	12.88	1	46
Jiangsu	36	46	1.28	0.51	1	3	41	27.39	13.99	2	46
Jiangxi	34	48	1.41	0.74	1	4	42	29.26	13.75	1	48
Jilin	32	42	1.31	0.69	1	4	36	22.89	11.65	1	42
Liaoning	42	61	1.86	1.20	1	6	43	40.95	25.52	1	78
Ningxia	28	34	1.21	0.50	1	3	42	19.79	10.89	1	34
Qinghai	30	40	1.33	0.61	1	3	40	22.73	12.46	1	40
Shaanxi	30	38	1.27	0.64	1	4	41	23.98	10.01	2	38
Shandong	37	75	2.03	1.42	1	8	41	47.85	21.45	2	75
Shanghai	41	80	1.95	1.09	1	5	44	47.36	25.70	1	80
Shanxi	29	42	1.45	0.69	1	3	42	24.52	12.50	1	42
Sichuan	39	59	1.51	0.88	1	5	43	36.72	17.07	1	59
Tianjin	34	81	2.38	1.61	1	6	43	52.26	25.87	1	81
Tibet	21	25	1.19	0.40	1	2	32	13.19	7.16	1	25
Xinjiang	31	35	1.13	0.34	1	2	41	20.29	10.21	1	35
Yunnan	36	67	1.86	0.99	1	5	42	39.67	20.00	1	67
Zhejiang	35	55	1.57	0.81	1	3	47	31.23	17.49	1	55
Total	1064	1849	1.74	1.58	1	17	1304	37.31	35.85	1	247

Notes: 1. N=sample size; 2. Daily number of news reports is the total number of official news reports and releases in one day for every province; 3. Ccumulative number of news reports is the cumulative number of daily news reports and releases each day.

Table A2

The relation between the number of news reports and the spread of COVID-19

	Daily identified patients	Daily contacted population
Panel A: with Hubei
# news	0.22	0.31
	(0.47)	(0.61)
# new 2	0.02	-0.15
	(0.06)	(0.11)
N	1243	1243
R2	0.75	0.55
Panel B: without Hubei
# news	0.89***	1.12*
	(0.20)	(0.63)
# news2	-0.09**	-0.29*
	(0.04)	(0.15)
N	1183	1183
R2	0.71	0.48

Notes: 1. Standard errors in parentheses; 2. * p<0.1, ** p<0.05, *** p<0.01; 3. # news: number of news releases and reports

Table C1

Descriptions of the key variables

Variable	Description
Epidemic
Daily patients	The number of daily new cases of COVID-19 for each province.
Daily contacted population	The number of daily individuals who had been in close contact with COVID-19 patients in each province.
Information Openness
# news	The daily cumulated number of official news releases and reports n COVID-19 in each province.
Population Mobility
Index of population inflow, 2020	Daily index of population inflow for each province, which indicates the population inflowed from other provinces to the target province in 2020.
Index of population inflow, 2019	Daily index of population inflow for each province, which indicates the population inflowed from other provinces to the target province in 2019.
Index of inner population movement, 2020	Daily index of inner population movement for each province, which indicates the inner population movement for target province in 2020.
Index of inner population movement, 2019	Daily index of inner population movement for every province, which indicates the inner population movement for target province in 2020.
Controls
Wind level	The level of wind for each province (daily).
Rain	0=None, 1=Rain, 2=Snow (daily).
Temperature	The average temperature for each province (daily).
Population size (million)	The whole population size for each province.
Area (10 thousand KM2)	The whole area for each province in kilometer squared.

Table C2

Summary statistics of the key variables

Variable	Obs	Mean	SD	Min	Max
Epidemiological
Daily number of new cases	1,863	43.05	418.07	0.00	14840.00
Daily number of close contacts	1,863	384.65	1203.03	0.00	12900.00
Media coverage
# news (number of news reports and releases)	1,860	23.55	32.45	0.00	236.00
Population Mobility
Index of population inflow, 2020	1,891	3.17	1.72	0.30	6.96
Index of population inflow, 2019	1,891	4.21	0.83	1.47	6.15
Index of inner population movement, 2020	1,922	3.62	4.25	0.04	28.75
Index of inner population movement, 2019	1,922	5.63	5.23	0.08	50.61
Controls
Wind level	1,916	2.70	1.12	2.00	7.00
Rain	1,916	0.27	0.55	0.00	2.00
Temperature	1,916	4.07	8.66	−23.50	25.50
Population size (million)	1,860	46.79	27.65	3.44	113.46
Area (10 thousand KM2)	1,860	31.00	38.12	0.63	166.00

Table D6

The direct and mediating effects of media coverage across provinces, excluding Hubei

	T		T1		T2		T		T1		T2
	R	C	R	C	R	C	R	C	R	C	R	C
Panel A: 3 days lag
# news, within	0.92⁎⁎⁎	1.13⁎⁎⁎	−0.57	−0.54	1.30⁎⁎⁎	1.25⁎⁎⁎	0.67⁎⁎⁎	−0.37	−0.33	0.69	1.15⁎⁎⁎	0.60*
	(0.09)	(0.25)	(0.44)	(0.99)	(0.14)	(0.29)	(0.09)	(0.25)	(0.28)	(0.62)	(0.14)	(0.31)
# news2, within	−0.61⁎⁎⁎	−0.68⁎⁎⁎	0.49	0.69	−0.87⁎⁎⁎	−0.79⁎⁎⁎	−0.46⁎⁎⁎	0.31*	0.19	−0.37	−0.78⁎⁎⁎	−0.35*
	(0.06)	(0.17)	(0.31)	(0.69)	(0.09)	(0.19)	(0.06)	(0.17)	(0.19)	(0.43)	(0.09)	(0.21)
# news, across	0.00	−0.00	0.03	0.09*	0.01	−0.01	0.01	−0.01	−0.01	0.03	0.01	−0.01
	(0.00)	(0.01)	(0.02)	(0.05)	(0.01)	(0.01)	(0.00)	(0.01)	(0.01)	(0.03)	(0.01)	(0.01)
inner movement, within	0.05*	0.04	0.11	−0.18	0.00	0.01	0.04	−0.03	0.19⁎⁎⁎	−0.42⁎⁎⁎	−0.00	0.01
	(0.03)	(0.07)	(0.11)	(0.26)	(0.03)	(0.07)	(0.02)	(0.06)	(0.07)	(0.16)	(0.03)	(0.07)
population inflow, within	−0.02	−0.22⁎⁎⁎	0.09	0.09	−0.03	−0.20⁎⁎⁎	0.03	−0.18⁎⁎⁎	0.05	−0.11	−0.01	−0.18⁎⁎
	(0.03)	(0.07)	(0.10)	(0.22)	(0.04)	(0.07)	(0.02)	(0.07)	(0.06)	(0.14)	(0.04)	(0.07)
population inflow, across	−0.01	−0.03	0.01⁎⁎⁎	0.01⁎⁎⁎	−0.01	−0.02	−0.00	−0.02	0.00⁎⁎	−0.01*	−0.01	−0.01
	(0.01)	(0.02)	(0.00)	(0.01)	(0.01)	(0.02)	(0.01)	(0.02)	(0.00)	(0.00)	(0.01)	(0.02)
Close contacts							0.23⁎⁎⁎		0.43⁎⁎⁎		0.12⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.63⁎⁎⁎		2.17⁎⁎⁎		0.50⁎⁎⁎
								(0.07)		(0.04)		(0.09)
N	1132	1132	663	663	469	469	1132	1132	663	663	469	469
Panel B: 5 days lag
# news, within	0.85⁎⁎⁎	1.41⁎⁎⁎	0.55	0.54	1.01⁎⁎⁎	1.37⁎⁎⁎	0.51⁎⁎⁎	1.51⁎⁎⁎	0.31	2.01⁎⁎⁎	0.80⁎⁎⁎	0.57⁎⁎⁎
	(0.10)	(0.25)	(0.39)	(0.84)	(0.15)	(0.29)	(0.10)	(0.07)	(0.28)	(0.05)	(0.15)	(0.09)
# news2, within	−0.56⁎⁎⁎	−0.87⁎⁎⁎	−0.32	−0.08	−0.69⁎⁎⁎	−0.86⁎⁎⁎	−0.35⁎⁎⁎	0.12	−0.28	−0.56	−0.56⁎⁎⁎	0.79⁎⁎⁎
	(0.07)	(0.17)	(0.29)	(0.63)	(0.10)	(0.20)	(0.06)	(0.24)	(0.21)	(0.61)	(0.10)	(0.31)
# news, across	0.01	−0.02	−0.05⁎⁎⁎	−0.06⁎⁎⁎	0.01	−0.02	0.01*	−0.01	−0.02⁎⁎⁎	0.55	0.01	−0.47⁎⁎
	(0.01)	(0.02)	(0.01)	(0.02)	(0.01)	(0.01)	(0.01)	(0.16)	(0.01)	(0.45)	(0.01)	(0.21)
inner movement, within	0.04*	−0.01	0.18	−0.20	0.03	−0.01	0.05⁎⁎	−0.03*	0.27⁎⁎⁎	0.04⁎⁎⁎	0.03	−0.02
	(0.02)	(0.06)	(0.13)	(0.28)	(0.03)	(0.06)	(0.02)	(0.01)	(0.09)	(0.01)	(0.03)	(0.01)
population inflow, within	−0.05*	−0.23⁎⁎⁎	−0.21⁎⁎	−0.27	−0.04	−0.14⁎⁎	0.01	−0.08	−0.09	−0.56⁎⁎⁎	−0.02	−0.03
	(0.03)	(0.06)	(0.10)	(0.21)	(0.03)	(0.07)	(0.02)	(0.06)	(0.07)	(0.20)	(0.03)	(0.06)
population inflow, across	−0.00	−0.03	−0.00	−0.01*	−0.00	−0.05⁎⁎	0.01	−0.16⁎⁎⁎	0.00	0.16	0.00	−0.11*
	(0.01)	(0.02)	(0.00)	(0.00)	(0.01)	(0.02)	(0.01)	(0.06)	(0.00)	(0.15)	(0.01)	(0.07)
Close contacts							0.24⁎⁎⁎	−0.03	0.44⁎⁎⁎	−0.01	0.15⁎⁎⁎	−0.04⁎⁎
							(0.01)	(0.02)	(0.01)	(0.00)	(0.02)	(0.02)
New cases								1.51⁎⁎⁎		2.01⁎⁎⁎		0.57⁎⁎⁎
								(0.07)		(0.05)		(0.09)
N	1105	1105	658	658	447	447	1105	1105	658	658	447	447
Panel C: 7 days lag
# news, within	0.83⁎⁎⁎	0.92⁎⁎⁎	0.84	0.46	1.20⁎⁎⁎	1.00⁎⁎⁎	0.59⁎⁎⁎	−0.26	0.62	−0.98	1.04⁎⁎⁎	0.24
	(0.11)	(0.26)	(0.60)	(1.15)	(0.17)	(0.33)	(0.11)	(0.25)	(0.47)	(0.90)	(0.17)	(0.34)
# news2, within	−0.57⁎⁎⁎	−0.60⁎⁎⁎	−0.16	0.62	−0.84⁎⁎⁎	−0.67⁎⁎⁎	−0.41⁎⁎⁎	0.21	−0.46	0.90	−0.73⁎⁎⁎	−0.15
	(0.08)	(0.18)	(0.48)	(0.91)	(0.11)	(0.22)	(0.07)	(0.17)	(0.37)	(0.71)	(0.11)	(0.23)
# news, across	0.02	−0.04	0.56⁎⁎	0.75	0.01	−0.03	0.03*	−0.07*	0.21	−0.21	0.02	−0.04
	(0.02)	(0.04)	(0.25)	(0.48)	(0.02)	(0.04)	(0.02)	(0.04)	(0.20)	(0.38)	(0.02)	(0.04)
inner movement, within	0.07⁎⁎	0.12	0.11	−0.49	0.06*	0.10	0.04	0.02	0.34⁎⁎	−0.69⁎⁎⁎	0.05	0.06
	(0.03)	(0.07)	(0.17)	(0.33)	(0.04)	(0.07)	(0.03)	(0.07)	(0.14)	(0.26)	(0.04)	(0.07)
population inflow, within	−0.01	−0.32⁎⁎⁎	−0.00	−0.10	−0.05	−0.21*	0.07	−0.30⁎⁎⁎	0.04	−0.09	−0.02	−0.18
	(0.05)	(0.12)	(0.12)	(0.22)	(0.06)	(0.12)	(0.05)	(0.12)	(0.09)	(0.17)	(0.06)	(0.11)
population inflow, across	0.04	−0.11	−0.11⁎⁎	−0.15*	0.02	−0.09	0.07	−0.17*	−0.04	0.04	0.04	−0.11
	(0.05)	(0.11)	(0.05)	(0.09)	(0.05)	(0.09)	(0.04)	(0.10)	(0.04)	(0.07)	(0.05)	(0.09)
Close contacts							0.27⁎⁎⁎		0.47⁎⁎⁎		0.17⁎⁎⁎
							(0.01)		(0.01)		(0.02)
New cases								1.42⁎⁎⁎		1.72⁎⁎⁎		0.63⁎⁎⁎
								(0.06)		(0.05)		(0.09)
N	1075	1075	650	650	425	425	1075	1075	650	650	425	425
Province FE	√	√	√	√	√	√	√	√	√	√	√	√
Date FE	√	√	√	√	√	√	√	√	√	√	√	√
Controls	√	√	√	√	√	√	√	√	√	√	√	√

Notes: 1. Standard errors in parentheses; 2. * p<0.1

p<0.05

p<0.01; 3. T=the full sample, T=subsample with data from Jan 1 to Feb 5, T=subsample with data from Feb 6 to Feb 29; 4. R=New cases, C=Close contacts, #news: cumulative number of news releases and reports.